From the OpenAI edition of October 30, 2026
Independent analysis: OpenAI's threshold for referring chats to police has moved from "may refer" to "we notify" since 2025
YLEM · Public safety · United States, Canada, China (comparative)
- Publisher
- YLEM
- Original publication
- October 25, 2026
- Source retrieved
- 2026-10-31
What happened
YLEM compares three OpenAI statements (Aug. 2025, Feb. 26, 2026 letter to Canada's AI minister, Apr. 28, 2026) and finds the referral standard moved from "may refer" with a concrete plan to "we notify" when risk is imminent and credible, without a stated target or time. It notes the Tumbler Ridge shooting case and that no referral count had been published as of Oct. 25.
Why it matters
Police, prosecutors and schools increasingly receive AI-vendor referrals; understanding the threshold helps agencies set intake procedures and assess evidentiary weight.
Evidence and measured results
Side-by-side quotation of public OpenAI statements with cited sources; the Q3 count OpenAI published Oct. 30 (200+) now answers the transparency gap the piece raised.
Limitations and uncertainty
Small independent publisher; relies on secondary reporting; also covers Anthropic, which belongs to its own stream and is not analyzed here.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-10-31; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- Agencies and schools receive AI-company threat referrals with no defined intake or evaluation standard.
- Stakeholders
- Police and sheriff intelligence units
- School threat-assessment teams
- Prosecutors
- Records officers
- Discovery questions
- Has your agency received an AI-vendor referral?
- How are such tips logged and retained?
- Who evaluates credibility?
- Value hypothesis
- A defined intake process may make vendor tips actionable and defensible.
- Potential engagement
- A bounded workshop to draft an AI-vendor referral intake and retention procedure.
- Unsupported claims
- That vendor referrals are accurate
- That referral volume indicates prevented harm
Pre-sales engineering
Role takeaway
- Fit
- Public-safety and school-safety intake processes.
- Architecture and integration
- Tip intake into existing CAD/RMS or threat-assessment case tools; not described in source.
- Prerequisites
- Tip intake channel
- Records retention schedule
- Constraints
- Vendor-defined thresholds
- Cross-border cases
- Security
- Referral content may include sensitive personal data; restrict access and log handling.
- Proof of value
- Tabletop exercise with a mock AI-vendor referral
Delivery
Role takeaway
- Work
- Draft and exercise an intake procedure for AI-vendor referrals.
- Dependencies
- Legal review
- Records policy
- Ownership
- Agency intelligence or threat-assessment unit.
- Skills and adoption
- Train intake staff on evaluating context-poor chat excerpts.
- Governance checkpoints
- Legal review of evidentiary use
- Privacy review
- Acceptance criteria
- Written intake procedure approved (proposed)
- One tabletop exercise completed (proposed)
- Risks
- Over-reaction to ambiguous chats
- Records and privacy exposure
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
Limited relevance.
Governance
Who approves, reviews and stays accountable for outcomes?
Vendor-defined thresholds, not statute, govern U.S. referrals; agencies have no say in the standard.
Security and privacy
What data, permissions and controls need testing?
Human reviewers read flagged chats; private or diary-like writing can become a referral, a privacy concern for students and employees.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Limited relevance.
Procurement
What should contracts, pricing and exit terms secure?
Limited relevance.
Operating model
Which teams own the service once it runs?
Adds a new external tip source to public-safety operations.
Publication history
- 2026-10-30OpenAI · Issue 027 resources
Stable resource ID: ylem-ai-police-referral-threshold-shift-20261025